qla2xxx: Add irq affinity notification

Register to receive notification of when irq setting change
occured.

Signed-off-by: Quinn Tran <quinn.tran@qlogic.com>
Signed-off-by: Himanshu Madhani <himanshu.madhani@qlogic.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>

3 files changed, 93 insertions, 1 deletions

diff --git a/drivers/scsi/qla2xxx/qla_def.h b/drivers/scsi/qla2xxx/qla_def.h
index 216a944c1ca5..c4fc32e9aaf9 100644
--- a/drivers/scsi/qla2xxx/qla_def.h
+++ b/drivers/scsi/qla2xxx/qla_def.h
@@ -2714,11 +2714,16 @@ struct isp_operations {
 
 struct scsi_qla_host;
 
+
+#define QLA83XX_RSPQ_MSIX_ENTRY_NUMBER 1 /* refer to qla83xx_msix_entries */
+
 struct qla_msix_entry {
 	int have_irq;
 	uint32_t vector;
 	uint16_t entry;
 	struct rsp_que *rsp;
+	struct irq_affinity_notify irq_notify;
+	int cpuid;
 };
 
 #define	WATCH_INTERVAL		1       /* number of seconds */
@@ -2930,6 +2935,7 @@ struct qlt_hw_data {
 	spinlock_t q_full_lock;
 	uint32_t leak_exchg_thresh_hold;
 	spinlock_t sess_lock;
+	int rspq_vector_cpuid;
 };
 
 #define MAX_QFULL_CMDS_ALLOC	8192
diff --git a/drivers/scsi/qla2xxx/qla_isr.c b/drivers/scsi/qla2xxx/qla_isr.c
index ea7e8e8a065a..cf0fe8ec12c7 100644
--- a/drivers/scsi/qla2xxx/qla_isr.c
+++ b/drivers/scsi/qla2xxx/qla_isr.c
@@ -18,6 +18,10 @@ static void qla2x00_status_entry(scsi_qla_host_t *, struct rsp_que *, void *);
 static void qla2x00_status_cont_entry(struct rsp_que *, sts_cont_entry_t *);
 static void qla2x00_error_entry(scsi_qla_host_t *, struct rsp_que *,
 	sts_entry_t *);
+static void qla_irq_affinity_notify(struct irq_affinity_notify *,
+    const cpumask_t *);
+static void qla_irq_affinity_release(struct kref *);
+
 
 /**
  * qla2100_intr_handler() - Process interrupts for the ISP2100 and ISP2200.
@@ -2548,6 +2552,14 @@ void qla24xx_process_response_queue(struct scsi_qla_host *vha,
 	if (!vha->flags.online)
 		return;
 
+	if (rsp->msix->cpuid != smp_processor_id()) {
+		/* if kernel does not notify qla of IRQ's CPU change,
+		 * then set it here.
+		 */
+		rsp->msix->cpuid = smp_processor_id();
+		ha->tgt.rspq_vector_cpuid = rsp->msix->cpuid;
+	}
+
 	while (rsp->ring_ptr->signature != RESPONSE_PROCESSED) {
 		pkt = (struct sts_entry_24xx *)rsp->ring_ptr;
 
@@ -2979,8 +2991,11 @@ qla24xx_disable_msix(struct qla_hw_data *ha)
 
 	for (i = 0; i < ha->msix_count; i++) {
 		qentry = &ha->msix_entries[i];
-		if (qentry->have_irq)
+		if (qentry->have_irq) {
+			/* un-register irq cpu affinity notification */
+			irq_set_affinity_notifier(qentry->vector, NULL);
 			free_irq(qentry->vector, qentry->rsp);
+		}
 	}
 	pci_disable_msix(ha->pdev);
 	kfree(ha->msix_entries);
@@ -3043,6 +3058,9 @@ qla24xx_enable_msix(struct qla_hw_data *ha, struct rsp_que *rsp)
 		qentry->entry = entries[i].entry;
 		qentry->have_irq = 0;
 		qentry->rsp = NULL;
+		qentry->irq_notify.notify  = qla_irq_affinity_notify;
+		qentry->irq_notify.release = qla_irq_affinity_release;
+		qentry->cpuid = -1;
 	}
 
 	/* Enable MSI-X vectors for the base queue */
@@ -3061,6 +3079,18 @@ qla24xx_enable_msix(struct qla_hw_data *ha, struct rsp_que *rsp)
 		qentry->have_irq = 1;
 		qentry->rsp = rsp;
 		rsp->msix = qentry;
+
+		/* Register for CPU affinity notification. */
+		irq_set_affinity_notifier(qentry->vector, &qentry->irq_notify);
+
+		/* Schedule work (ie. trigger a notification) to read cpu
+		 * mask for this specific irq.
+		 * kref_get is required because
+		* irq_affinity_notify() will do
+		* kref_put().
+		*/
+		kref_get(&qentry->irq_notify.kref);
+		schedule_work(&qentry->irq_notify.work);
 	}
 
 	/*
@@ -3240,3 +3270,47 @@ int qla25xx_request_irq(struct rsp_que *rsp)
 	msix->rsp = rsp;
 	return ret;
 }
+
+
+/* irq_set_affinity/irqbalance will trigger notification of cpu mask update */
+static void qla_irq_affinity_notify(struct irq_affinity_notify *notify,
+	const cpumask_t *mask)
+{
+	struct qla_msix_entry *e =
+		container_of(notify, struct qla_msix_entry, irq_notify);
+	struct qla_hw_data *ha;
+	struct scsi_qla_host *base_vha;
+
+	/* user is recommended to set mask to just 1 cpu */
+	e->cpuid = cpumask_first(mask);
+
+	ha = e->rsp->hw;
+	base_vha = pci_get_drvdata(ha->pdev);
+
+	ql_dbg(ql_dbg_init, base_vha, 0xffff,
+	    "%s: host %ld : vector %d cpu %d \n", __func__,
+	    base_vha->host_no, e->vector, e->cpuid);
+
+	if (e->have_irq) {
+		if ((IS_QLA83XX(ha) || IS_QLA27XX(ha)) &&
+		    (e->entry == QLA83XX_RSPQ_MSIX_ENTRY_NUMBER)) {
+			ha->tgt.rspq_vector_cpuid = e->cpuid;
+			ql_dbg(ql_dbg_init, base_vha, 0xffff,
+			    "%s: host%ld: rspq vector %d cpu %d  runtime change\n",
+			    __func__, base_vha->host_no, e->vector, e->cpuid);
+		}
+	}
+}
+
+void qla_irq_affinity_release(struct kref *ref)
+{
+	struct irq_affinity_notify *notify =
+		container_of(ref, struct irq_affinity_notify, kref);
+	struct qla_msix_entry *e =
+		container_of(notify, struct qla_msix_entry, irq_notify);
+	struct scsi_qla_host *base_vha = pci_get_drvdata(e->rsp->hw->pdev);
+
+	ql_dbg(ql_dbg_init, base_vha, 0xffff,
+	    "%s: host%ld: vector %d cpu %d \n", __func__,
+	    base_vha->host_no, e->vector, e->cpuid);
+}
diff --git a/drivers/scsi/qla2xxx/qla_target.c b/drivers/scsi/qla2xxx/qla_target.c
index 6136987df164..9a4aed0e8241 100644
--- a/drivers/scsi/qla2xxx/qla_target.c
+++ b/drivers/scsi/qla2xxx/qla_target.c
@@ -6232,6 +6232,7 @@ qlt_enable_vha(struct scsi_qla_host *vha)
 	struct qla_tgt *tgt = vha->vha_tgt.qla_tgt;
 	unsigned long flags;
 	scsi_qla_host_t *base_vha = pci_get_drvdata(ha->pdev);
+	int rspq_ent = QLA83XX_RSPQ_MSIX_ENTRY_NUMBER;
 
 	if (!tgt) {
 		ql_dbg(ql_dbg_tgt, vha, 0xe069,
@@ -6250,6 +6251,17 @@ qlt_enable_vha(struct scsi_qla_host *vha)
 		qla24xx_disable_vp(vha);
 		qla24xx_enable_vp(vha);
 	} else {
+		if (ha->msix_entries) {
+			ql_dbg(ql_dbg_tgt, vha, 0xffff,
+			    "%s: host%ld : vector %d cpu %d\n",
+			    __func__, vha->host_no,
+			    ha->msix_entries[rspq_ent].vector,
+			    ha->msix_entries[rspq_ent].cpuid);
+
+			ha->tgt.rspq_vector_cpuid =
+			    ha->msix_entries[rspq_ent].cpuid;
+		}
+
 		set_bit(ISP_ABORT_NEEDED, &base_vha->dpc_flags);
 		qla2xxx_wake_dpc(base_vha);
 		qla2x00_wait_for_hba_online(base_vha);